Assessment of computational model updating procedures with regard to model validation

Abstract

In this paper two different computational model updating (CMU) methods are compared. Both procedures make use of the inverse sensitivity approach. The residuals of the first method are formed by eigenvalue and mode shape differences whereas the residuals of the second method include eigenvalue and frequency response errors. Computational model updating of physical mass, stiffness and geometric parameters is possible with the first technique. The second method additionally allows updating of viscous modal damping parameters. This second method uses directly measured frequency response data taken only at the resonance peaks. Since both methods allow to handle incomplete test data vectors, that means that the number of measured degrees of freedom (dof) is much less than the dof no. of the mathematical model, they are able to deal with large order aerospace finite element (FE) models. Both procedures were used to investigate the prediction capability of an updated analytical model of a benchmark test structure with respect to the test data of a modified structure. The results of both techniques are compared, advantages and disadvantages are pointed out and an assessment with respect to the operational validity is given.